Validation of methods performance for routine biochemistry analytes

Special issue:
Original
articleResponsible writing in science
Validation of methods performance for routine biochemistry analytes at Cobas
6000 analyzer series module c501
Vesna Supak Smolcic1*, Lidija Bilic-Zulle1,2, Elizabeta Fisic1
1Clinical
Institute of Laboratory Diagnostics, Rijeka Clinical Hospital Center, Rijeka, Croatia
of Medical Informatics, Rijeka University School of Medicine, Rijeka, Croatia
2Depar tment
*Corresponding author: vesnasupak@gmail.com
Abstract
Introduction: Cobas 6000 (Roche, Germany) is biochemistry analyzer for spectrophotometric, immunoturbidimetric and ion-selective determination of biochemical analytes. Hereby we present analytical validation with emphasis on method performance judgment for routine operation.
Materials and methods: Validation was made for 30 analytes (metabolites, enzymes, trace elements, specific proteins and electrolytes). Research
included determination of within-run (N = 20) and between-run imprecision (N = 30), inaccuracy (N = 30) and method comparison with routine
analyzer (Beckman Coulter AU640) (N = 50). For validation of complete analytical process we calculated total error (TE). Results were judged according to quality specification criteria given by European Working Group.
Results: Within-run imprecision CVs were all below 5% except for cholesterol, triglycerides, IgA and IgM. Between-run CVs for all analytes were
below 10%. Analytes that did not meet the required specifications for imprecision were: total protein, albumin, calcium, sodium, chloride, immunoglobulins and HDL cholesterol. Analytes that did not fulfill requirements for inaccuracy were: total protein, calcium, sodium and chloride. Analytes
that deviated from quality specifications for total error were: total protein, albumin, calcium, sodium, chloride and IgM. Passing-Bablok regression
analysis provided linear equation and 95% confidence interval for intercept and slope. Complete accordance with routine analyzer Beckman Coulter
AU640 showed small number of analytes. Other analytes showed small proportional and/or small constant difference and therefore need to be adjusted for routine operation.
Conclusions: Regarding low CV values, tested analyzer has satisfactory accuracy and precision and is extremely stable. Except for analytes that are
coherent on both analyzers, some analytes require adjustments of slope and intercept for complete accordance.
Key words: validation study; chemistry techniques, analytical; Cobas 6000 analyzer; total error; Passing and Bablock regression
Received: February 15, 2011
Accepted: April 12, 2011
Introduction
It is well documented that routine laboratory work
is not error free and efforts have been made to improve whole laboratory testing cycle and reduce
errors (1). Although the analytical phase is the least
prone to errors, it still has room for improvement
(2). For that purpose many guidelines, protocols
and specifications for validation of analytical systems are available. They all are quite similar, but some discrepancies exist usually in acceptance criteria depending on source (3-5).
Quality parameters, inaccuracy and imprecision
are basic parameters of methods performance vaBiochemia Medica 2011;21(2):182-90
182
lidation as measure of systematic and random error. Those parameters are presented by coefficient
of variation (CV) and bias, but also can be used for
calculation of total error (TE) (6). Combining effects of systemic and random error in form of total error makes validation results of complete analytical
process more evident.
The aim of our study was to validate methods performance of biochemistry analyzer Cobas 6000
analyzer series module c501 in context of our operating conditions.
Supak Smolcic V. et al.
Materials and methods
Analyzer and methods description
Cobas 6000 analyzer series is composed of several
units (modules) that can be combined in seven different analytical systems with only one load point.
Various analyzer compositions are completely adjusted to user’s needs for clinical chemistry and
immunochemistry analyses with high throughput
capacity (7). We validated clinical chemistry analytical unit c501 for photometric and ion-selective
electrode (ISE) measurement. For validation of
analyzer in routine laboratory work we tested analytes representing metabolites (glucose, urea, creatinine, total and direct bilirubin, uric acid, triglycerides, total cholesterol, HDL (high-denstiy lipoprotein) and LDL (low-denstiy lipoprotein) cholesterol), enzymes (amylase, alkaline phosphatase (AP),
alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyltransferase
(GGT), creatin kinase (CK), lactate dehydrogenase
(LDH) and lypase), electrolytes (sodium, potassium,
chloride, magnesium, inorganic phosphorus and
calcium), proteins (total protein, albumin, immunoglobulins (Ig) G, A and M) and trace elements
(iron). The principle of each assay is listed in table
1. Reagents are ready to use and packed in closed
cassettes what makes reagent handling completely automated. According to the manufacturer, Cobas 6000 analyzer module c501 performs 600 analyses per hour and sets 130 different applications
in various body fluids.
Reagents
All measurements were performed with single lot
cassettes except for IgM where analysis of between-day imprecision and therefore inaccuracy was
performed with two different reagent lots. Reagent preparation and setting was made according to
Roche diagnostics recommendation. All calibrations were performed with commercial calibrators
and according to manufacturer’s instructions (7).
Calibrations were performed when quality control
measurement was not satisfactory or after reagent
lot change (IgM). All reagents, calibrators and controls used were provided by Roche diagnostics,
Mannheim, Germany.
Methods performance validation at Cobas c501
Samples
We tested 30 analytes in blood serum as listed in
table 1. Commercial control samples were used for
determination of between-day imprecision and
for calculation of inaccuracy (bias) (Precinorm U lot
17959600, Precipath U lot 17628700, Precinorm
Protein lot 18234300, Precipath Protein lot
18234400, Precinorm Lipid lot 18064700 and Precipath H/LDL lot 18174000, Roche diagnostics, Mannheim, Germany). For determination of within-day
imprecision and for method comparison fresh residual patients’ samples from daily routine were
used. Samples were non-randomly selected to fulfill criteria for broad value range and proper distribution for methods comparison. Only non-hemolytic and non-lipemic sera were used.
Imprecision and inaccuracy
Two patients’ samples, each with different decision
level (analyte concentration) were used for obtaining within-day imprecision. Each sample was
measured 20 times in series, and mean, standard
deviation and coefficient of variation (CV) were
calculated for each level of repeated measures.
Mean value of both coefficients of variation was
considered as final coefficient of variation for within-day imprecision (CVwd). Between-day imprecision was calculated based on analyses of commercial control samples, one with normal and one with pathological values (Roche, Mannheim, Germany) over 30 days period. Mean of coefficients of
variation for both measured controls levels was
considered as final coefficient of variation of between-day imprecision (CVbd). Between-day imprecision (CVbd) was considered to be the measure
for random analytical error (8).
Results of commercial control material analysis for
between-day imprecision measurements were also used for inaccuracy calculations (bias). We calculated percentage of bias according to equation:
Bias (%) = ((Mean value – Target value)/
Target value) × 100%.
Inaccuracy was considered to be the measure for
systemic analytical error (8). Target values for selected analytes in commercial controls were provided by the reagent manufacturer (Roche, Mannheim, Germany).
Biochemia Medica 2011;21(2):182-90
183
Supak Smolcic V. et al.
Methods performance validation at Cobas c501
TABLE 1. Methods of determination for analytes on Cobas 6000 analyzer series module c501 and Beckman Coulter AU640 (Linearity
ranges are stated by the manufacturer)
Analyte (unit)
Glucose (mmol/L)
Urea (mmol/L)
Linearity
Cobas c501
Method used on Cobas 6000
analyzer series modul c501
Linearity Beckman
Coulter AU640
Method used on Beckman
Coulter AU640
0.11–41.6
Hexokinase
0.6–45.0
Hexokinase
0.5–40
Urease-GLDH
0.8–50
Urease-GLDH
Creatinine (µmol/L)
15–2200
Kinetic Jaffé
18–2200
Kinetic Jaffé
Total bilirubin (µmol/L)
1.7–650
DPD
0–513
DPD
Direct bilirubin (µmol/L)
2–430
Diazo method
0–171
Diazo method
AMY (U/L)
3–1500
IFCC (enzymatic with E-G7PNP)
10–1500
IFCC (enzymatic with E-G7PNP)
ALP (U/L)
5–1200
IFCC
5–1500
IFCC
LDH (U/L)
20–1200
IFCC
3–1000
IFCC
GGT (U/L)
3–1200
IFCC
5–1200
IFCC
AST (U/L)
5–700
IFCC without pyridoxal phosphate
3–1000
IFCC without pyridoxal
phosphate
ALT (U/L)
5–700
IFCC without pyridoxal phosphate
3–500
IFCC without pyridoxal
phosphate
CK (U/L)
7–2000
IFCC
10–2000
IFCC
Lipase (U/L)
3–300
Enzymatic colorimetric assay
dilauril -glicerol ester
3–600
Kinetic colorimetric assay
GPO-PAP
11.9–1487
Uricase–PAP
89–1785
Uricase-PAP
Triglycerides (mmol/L)
0.1–10.0
GPO-PAP with 4-aminophenazone
0.1–11.3
GPO-PAP with 4-aminoantipyrine
Cholesterol (mmol/L)
0.1–20.7
CHOD-PAP
0.5–18.0
CHOD-PAP
Total protein (g/L)
2.0–120
Biuret
30–120
Biuret
Albumin (g/L)
2–60
Colorimetric with BCG
15–60
Colorimetric with BCG
Iron (µmol/L)
0.9–179
Colorimetric assay with Ferrozine
2–179
Colorimetric with TPTZ
Magnesium (mmol/L)
0.1–2.5
Colorimetric method with
chlorophosphonazo III
0.2–3.3
Colorimetric with xilidil blue
Inorganic phosphorous
(mmol/L)
0.1–6.46
Molybdate UV method
0.32–6.40
Molybdate UV method
Calcium (mmol/L)
0.1–5.0
Colorimetric method according
to Schwarzenbach with
o-cresolphthalein
1.0–5.0
Colorimetric method with
arsenazo III
Na+ (mmol/L)
80–180
Indirect potenciometry
50–200
Indirect potenciometry
K+ (mmol/L)
1.5–10.0
Indirect potenciometry
1.0–10.0
Indirect potenciometry
Cl+ (mmol/L)
60–140
Indirect potenciometry
50–200
Indirect potenciometry
IgA (g/L)
0.5–8.0
Immunoturbidimetric assay
0.1–7.0
Immunoturbidimetric assay
IgG (g/L)
3.0–50.0
Immunoturbidimetric assay
0.75–30.0
Immunoturbidimetric assay
IgM (g/L)
0.25–6.5
Immunoturbidimetric assay
0.2–5.0
Immunoturbidimetric assay
HDL (mmol/L)
0.08–3.1
Homogeneous enzymatic
colorimetric assay with
4-aminoantipyrine
0.05–4.65
Homogeneous enzymatic
colorimetric assay with
4-aminoantipyrine
LDL (mmol/L)
0.1–14.2
Homogeneous enzymatic
colorimetric assay with
4-aminoantipyrine
0.26–10.3
Homogeneous enzymatic
colorimetric assay with
4-aminoantipyrine
Uric acid (µmol/L)
Biochemia Medica 2011;21(2):182-90
184
Supak Smolcic V. et al.
Results on imprecision and inaccuracy were evaluated according to quality specifications recommended by European Working Group (9).
Total error
Data on inaccuracy (bias) and imprecision (CVbd)
were used for calculation of total error (TE) according to equation (6):
TE = Bias + Z × CVbd.
Total error (TE) represents overall error that occurs
as combined effect of random error (imprecision)
and systemic error (inaccuracy) in analytical measurement. Factor Z is multiplier that sets confidence level. Recommendations for Z value are not
completely harmonized and Z can range from 2 to
6, the higher the value the stricter the rule (6). We
calculated total error using Z = 2 and compared
our results with recommended quality specifications e.g. total allowable error (TEa) according to
European Working Group (9).
Method comparison
We compared results of analysis obtained on Cobas 6000 analyzer series module c501 with results
obtained on currently used routine biochemistry
analyzer, Beckman Coulter AU640 (Beckman Coulter, USA) for all investigated methods (Table 1). For
each analyte we performed measurements of 50
patients’ samples in period of one month. Samples
were analyzed on both analyzers at the same day
so there was no need for sample storage. We performed Passing and Bablok regression analysis for
data of all analytes. Regression analysis provided
linear equation for each tested analyte as well as
95% confidence interval (95% CI) for intercept and
slope. Methods were considered harmonized if
95% CI for intercept included value zero and 95%
CI for slope included value one. If 95% CI did not
include listed values, proportional and constant
error could be identified (10).
Statistical analysis
All measured data were entered in relation tables
using MS Excel (Microsoft Corporation, USA). Computation and statistical analysis were made using
MedCalc statistical software (MedCalc, Mariakerke,
Belgium, licence of Department of Medical Infor-
Methods performance validation at Cobas c501
matics, Rijeka University School of Medicine, Rijeka, Croatia).
Whole validation experiment was performed at
Clinical Institute of Laboratory Diagnostics, Rijeka
Clinical Hospital Center, from February to April
2009.
Results
Results of validation of method performance are
presented in table 2 as mean of within-run CV, total imprecision, total inaccuracy and total error for
all analytes. Quality specifications for imprecision,
inaccuracy and total error for each analyte are presented in order to compare obtained with recommended values (9).
Coefficients of variation (CVwd) for within-day imprecision for majority of tested analytes were below 5% except for cholesterol, triglycerides, IgA
and IgM that were 7.4%, 7.7%, 5.4% and 12.3% respectively (Table 2). Coefficients of variation for between-day imprecision (CVbd) for all analytes were
below 10% but when compared to quality specifications, total protein, albumin, calcium, sodium,
chloride, IgG, IgA, IgM and HDL cholesterol are
higher than recommended. When comparing total inaccuracy (bias) with quality recommendations, total protein, calcium, sodium and chloride are
not in accordance with recommended specifications. Results for total error revealed that total protein, albumin, calcium, sodium, chloride and IgM
have higher total error value than recommended
(Table 2).
Method comparison study yielded coefficient of
correlation r > 0.98 for majority of analytes except
for HDL cholesterol (r = 0.97), sodium (r = 0.97),
magnesium and chloride (r = 0.95 for both) (Table
3). Linear equations for all tested analytes as well
as 95% CI for intercept and slope, calculated according to Passing and Bablok regression analysis are
presented in table 3. Full accordance between two
methods were achieved for albumin, amylase,
chloride, HDL cholesterol, LDH, potassium, sodium
and urea, because 95% CI for regression line intercept includes value zero and 95% CI for slope includes value one. Method comparison regression
analysis for AST, ALT, IgM, iron, LDL cholesterol, lipase, magnesium, total protein and triglycerides
Biochemia Medica 2011;21(2):182-90
185
Supak Smolcic V. et al.
Methods performance validation at Cobas c501
TABLE 2. Within-run imprecision data (N = 20) and between-day imprecision data (N = 30) for Cobas 6000 analyzer series module
c501, Total imprecision is mean CV value of between-day coefficient of variation given for two levels of control material, Total inaccuracy as deviation percentage of mean measured value vs. target value (bias = ((mean value-target value)/target value) × 100%), Total
error (TE) as quality characteristics of validated method according to equation TE = bias + 2 x CV (6), the net or combined effect of
random (total imprecision) and systematic (inaccuracy) errors. All measured and calculated values are compared with given quality
specifications (9)
Analyte
Glucose (mmol/L)
Total imprecision
Total inaccuracy
Mean of
within-run
Mean of
Cobas c501
Quality
Quality
coefficient between-day
bias vs.
spe
ci
fi
ca
tion
spe
cification
of variation coefficient of
Control
(%)
(%)
(%)
variation (%)
material (%)
3.2
1.5
2.9
1.7
2.2
Total error (TE)
Total
error
(%)
Quality
specification
(%)
4.7
6.9
Urea (mmol/L)
4.0
2.1
6.2
4.4
5.5
8.6
15.7
Creatinine (µmol/L)
4.5
2.4
2.7
1.5
3.8
6.3
8.2
Total bilirubin (µmol/L)
1.6
2.6
11.9
2.6
11.4
7.8
31.1
Direct bilirubin (µmol/L)
2.2
4.8
18.4
6.0
14.2
15.6
44.5
AMY (U/L)
0.7
2.1
4.4
2.1
7.4
6.3
14.6
ALP (U/L)
4.2
2.6
3.2
4.9
6.4
10.1
11.7
LDH (U/L)
3.7
1.7
4.3
1.3
4.3
4.7
11.4
GGT (U/L)
1.1
1.7
6.9
4.0
10.8
7.4
22.2
AST (U/L)
0.6
2.3
6.0
1.9
5.4
6.5
15.2
ALT (U/L)
0.6
1.6
12.2
3.9
12.0
7.1
32.1
CK (U/L)
0.7
1.4
11.4
1.2
11.5
4.0
30.3
Lipase (U/L)
1.0
2.1
11.6
3.8
10.1
8.0
29.1
Uric acid (µmol/L)
4.3
1.7
4.5
4.5
4.9
7.9
12.4
Triglycerides (mmol/L)
7.7
1.9
10.5
0.9
10.7
4.7
27.9
Cholesterol (mmol/L)
7.4
1.5
2.7
3.3
4.0
6.3
8.5
Total protein (g/L)
1.0
2.1
1.4
1.6
1.2
5.8
3.4
Albumin (g/L)
1.8
1.9
1.6
0.8
1.3
4.6
3.9
Iron (µmol/L)
0.5
2.5
13.3
3.5
8.8
8.5
30.7
Magnesium (mmol/L)
1.2
1.4
1.8
0.5
1.8
3.3
4.8
Inorganic phosphorous
(mmol/L)
0.8
1.4
4.3
0.7
3.2
3.5
10.2
Calcium (mmol/L)
0.7
2.1
1.0
2.1
0.8
6.3
2.4
Sodium (mmol/L)
0.4
1.4
0.4
0.6
0.3
3.4
0.9
Potasium (mmol/L)
0.5
1.8
2.4
0.5
1.8
4.1
5.8
Chloride (mmol/L)
0.6
1.4
0.6
1.9
0.5
4.7
1.5
IgA (g/L)
5.4
2.9
2.7
3.9
9.1
9.7
13.5
IgG (g/L)
3.9
3.0
2.3
1.2
4.3
7.2
8.0
IgM (g/L)
12.3
6.9
3.0
5.5
11.9
19.3
16.8
HDL (mmol/L)
1.1
3.4
2.8
1.2
6.9
8.0
11.5
LDL (mmol/L)
2.0
3.4
4.2
1.7
6.8
8.5
13.6
Biochemia Medica 2011;21(2):182-90
186
Supak Smolcic V. et al.
Methods performance validation at Cobas c501
TABLE 3. Results of method comparison between Cobas 6000 analyser series c501 and Beckman Coulter AU640. Intercepts and slopes with 95% CI according to Passing and Bablock regression analysis and correlation coefficient are presented
Analyte
Intercept
95% CI for
intercept
Slope
95% CI for
slope
Correlation Range tested
coefficient
(Cobas)
In accordance with routine method*
Albumin
0.97
-0.39–2.4
1.02
0.98–1.06
0.99
15.9–54.2
Amylase
0.00
-0.25–0.00
1.00
1.00–1.01
0.99
17–356
Chloride
-4.00
-14.72–5.58
1.00
0.92–1.11
0.95
79–124
HDL
0.00
0.00–0.19
1.00
0.89–1.00
0.97
0.7–2.5
LDH
0.14
-5.52–6.32
0.99
0.96–1.02
0.99
119–488
Potassium
-0.12
-0.33–0.10
1.04
1.00–1.08
0.99
2.9–6.2
Sodium
1.00
-4.91–9.56
1.00
0.94–1.04
0.97
117–160
-0.10
-0.25–(-0.01)
1.00
0.99–1.03
0.99
1.0–24.8
ALT
1.00
0.75–1.00
1.00
1.00–1.01
0.99
1–124
AST
-0.85
-1.00–(-0.25)
0.99
0.97–1.00
0.99
9–198
IgM
-0.10
-0.10–(-0.10)
1.00
1.00–1.00
0.99
0.1–19.3
Iron
1.00
1.00–1.00
1.00
1.00–1.00
0.99
2–62
LDL
-0.16
-0.37–(-0.1)
1.02
1.00–1.09
0.98
1.0–6.3
Lipase
4.38
1.35–6.52
1.02
0.97–1.06
0.99
6.9–821.2
Magnesium
0.12
0.08–0.21
0.94
0.83–1.00
0.95
0.4–1.4
Total protein
1.00
1.00–5.42
1.00
0.93–1.00
0.99
38–115
Triglycerides
0.10
0.10–0.14
1.00
0.95–1.00
0.99
0.5–4.9
ALP
1.92
-0.73–3.3
0.88
0.87–0.90
0.99
26–609
Calcium
0.01
-0.09–0.15
0.89
0.84–0.94
0.98
0.9–2.7
Cholesterol
-0.02
-0.02–0.04
0.92
0.90–0.94
0.99
2.0–8.0
CK
0.07
-1.00–0.96
0.97
0.96–0.98
0.99
12–1742
IgG
0.20
-0.22–0.52
0.87
0.83–0.90
0.99
0.2–27.2
Uric acid
0.58
-3.90–6.90
0.91
0.89–0.93
0.99
84–680
-10.49
-12.57–(-8.41)
1.05
1.03–1.07
0.99
32–256
Direct bilirubin
-1.00
-1.33–(-0.25)
1.50
1.37–1.67
0.99
1–93
GGT
-1.30
-1.58–(-1.07)
0.96
0.96–0.97
0.99
4–501
Glucose
-0.15
-0.25–(-0.01)
1.07
1.05–1.09
0.99
2.7–14.7
IgA
-0.05
-0.14–(-0.07)
0.84
0.81–0.92
0.99
0.3–10.6
Inorganic phosphorous
0.05
0.01–0.08
0.91
0.88–0.95
0.99
0.6–1.9
Total bilirubin
-1.14
-1.88–(-0.85)
0.86
0.84–0.88
0.99
3–165
Urea
Constant
difference†
Propor tional difference‡
Both constant and propor tional
Creatinine
difference§
* Intercept
CI includes zero as value and slope CI includes one as value
† Intercept CI does not include zero as value and slope CI includes one as value
‡ Intercept CI includes zero as value and slope CI does not include one as value
§ Intercept CI does not include zero as value and slope CI does not include one as value
Biochemia Medica 2011;21(2):182-90
187
Supak Smolcic V. et al.
revealed constant difference (intercept 95% CI
does not include value zero), and for ALP, calcium,
cholesterol, CK, uric acid and IgG proportional difference (slope 95% CI does not include value one).
Both, constant and proportional differences, were
found for glucose, creatinine, direct bilirubin, total
bilirubin, GGT, inorganic phosphorous and IgA.
Discussion
Cobas 6000 analyzer series modul c501 is stable
analytical system for routine laboratory work. That
is supported by results for within-run coefficient of
variation that is generally lower than 5% (except
for triglycerides, cholesterol, IgA and IgM) and for
between-day coefficient of variation lower than
10% for all tested analytes (Table 2). Our results on
validation are consistent with those published by
van Gammeren et al. (11). Although results are presented in different manner, data for total imprecision can be compared, and no significant difference between obtained results is noted (11).
As we focused mainly on methods performance
rather than instrument validation our results were
observed in detail according to recommended
quality specifications for each method. There are
no uniform requirements for quality specification
accepted globally (12-15). The choice of quality
specification depends on laboratory management
and we accepted European Working Group recommendations (9).
When comparing values of CV and bias obtained
in our study, with quality specifications recommended by European Working Group, results for
some analytes do not agree with recommendations (9) (Table 2). For total protein, albumin, calcium,
sodium, chloride, IgA, IgG and IgM, and HDL-cholesterol, CV is higher than recommended, and for
total protein, calcium, sodium and chloride bias is
also higher than recommended by quality specification (Table 2). Evaluation of quality of methods
through CV and bias separately can be complex
and inconclusive. CV and bias are performance
characteristics of measurement procedures, but
better quality indicator is total error that describes
maximum error that can occur as consequence of
imprecision and inaccuracy of particular analyte
measurement (6).
Biochemia Medica 2011;21(2):182-90
188
Methods performance validation at Cobas c501
When we observe our results in the terms of total
error and compare them with European quality
specification for allowable total error, slightly different conclusions can be derived. Total protein, albumin, calcium, sodium, chloride and IgM did not
meet quality specification criteria for total error
(Table 2). Analytes that did not meet quality specifications for CV and for bias, failed to meet quality
specification for total error as well (total protein,
calcium, sodium and chloride). However, IgA, IgG
and HDL-cholesterol failed to meet CV criteria but
were acceptable according to total error quality
specification. Although albumin and IgM have acceptable bias, discrepancies from recommended
CV value were high enough to influence total errors that were higher than recommended quality
specification. Recommended quality specifications for allowable total error by European Working
group are more demanding than, for example US
CLIA (Clinical Laboratory Improvement Amendments) recommendations (16,17). If we used CLIA
recommendation, all analytes that did not meet
European quality specifications with the exception
of IgM would meet CLIA criteria for allowable total
error. The possible reason for poor performance of
IgM may lay in two reagent lots that have been
used during the validation experiment. Both reagents were calibrated with commercial calibrators
according to manufacturer’s recommendations
and passed commercial quality control criteria (7).
Factor that strongly influences total error value is
Z-value used for calculation of TE. Total errors in
our study were calculated using factor Z = 2, however recommended quality specifications for
allowable total error by European working group
were given using factor Z = 1.65. Considering that
higher Z value means stricter rule, we used higher
value than the one recommended by European
working group and thus made our quality parameters more demanding. The choice of Z value
depends on quality policy of laboratory (6). Unstable environment such as routine laboratory is
controlled every day with commercial control samples so it allows consideration of factor Z = 2 as satisfactory. Higher stability of analytical systems
would allow fewer controls (6). Methods that do
not fulfill quality specification criteria should not
be declared unsatisfactory but as methods with
Supak Smolcic V. et al.
lower stability. Those methods should be controlled more often and followed with more attention
because of higher possible maximum error in method performance (6,18).
Method comparison is inevitable procedure when
new analytical system is introduced in routine laboratory work. Regardless of results of method
performance validation based on CV, bias and TE,
new methods have to preserve continuum in medical decision process so it is important that they
do not differ significantly from those currently
used (18). We used Passing and Bablock regression
analysis for method comparison evaluation even
though some authors recommend use of difference plots (Bland and Altman analysis) to present data from method comparison study because they
provide more information on random errors (10,19).
Our aim was rather to estimate systematic errors,
both constant and proportional in order to harmonize tested analyzers.
According to our results of method comparison
experiment, methods for albumin, amylase, chloride, HDL-cholesterol, LDH, potassium, sodium and
urea were in full concordance with methods of
existed automated analyzer (Beckman Coulter
AU640, Beckman Coulter, USA). Methods that do
not completely fulfill criteria for full concordance
(Table 3) generally have small constant and proportional difference that can be compensated by
slight adjustment of slope and intercept except for
direct bilirubin. That method has significant proportional difference (slope = 1.50) and both instruments cannot be used simultaneously in routine laboratory work for measurement of direct bilirubin.
The limitation of study is that the main question
remains, whether those methods that did not fulfill strict performance criteria influence clinical quality requirements which are basis for identifying
Methods performance validation at Cobas c501
medically important changes (20). One of the limitations is the exclusion of hemolytic and lipemic
samples from method validation as we did not include interference analysis in our study. Another limitation is the usage of control material provided
by the manufacturer of the tested analyzer as other control materials from independent manufacturer were not available. Evaluation of each method was performed using single lot reagent with
the exception of IgM. Although this can be considered as limitation of the study, validation is a
compromise of cost and risk (5). Our study only
accessed analytical performance of the specified
methods. Validation of analytical performance
should never be mixed with clinical quality requirements. There are different protocols for clinical
quality requirement evaluation that were not included in our study (21-23).
To conclude, Cobas 6000 analyzer series module
c501 provides stable analysis for wide range of
analytes. Analytes that do not fulfill quality specification criteria (total protein, albumin, calcium, sodium, chloride and IgM) require more frequent
quality control protocol which includes several runs of control material in series. Also, there is need
for continuous monitoring of bias and CV in routine laboratory work and stricter rules for allowable
CV and bias (18). Additional comparison study for
direct bilirubin method adjustment needs to be
performed in order to harmonize it with existent
instrument.
Acknowledgements
We thank Roche Diagnostics for reagent supply for
validation study.
Potential Conflicts of Interest: Roche Diagnostics
supported us with reagents.
References
1. Bonini PA, Plebani M, Ceriotti F, Ruboli FF. Errors in laboratory medicine. Clin Chem 2002;48:691-8.
2. Lippi G, Plebani M, Simundic AM. Quality in laboratory diagnostics: from theory to practice. Biochem Med 2010;20):126-30.
3. Westgard JO, Method validation – The real world applications. Available at: http://www.westgard.com/qcapp14.
htm. Accessed February 12th, 2011.
4. Fraser CG, Hytloft Petersen P. Analytical performance characteristics should be judged against objective quality specifications. Clin Chem 1999;45:321-3.
5. Gašljević V. Method validation and measurement uncertainty. Biochem Med 2010;20:57-63.
6. Westgard JO. The meaning and Application of total error.
Available at: http://www.westgard.com/essay111.htm. Accessed February 12th, 2011
Biochemia Medica 2011;21(2):182-90
189
Supak Smolcic V. et al.
7. Cobas 6000 analyzer series operator’s manual, Roche, Mannheim, Germany.
8. Westgard JO, Carey RN, Wold S. Criteria for judging precision and accuracy in method development and evaluation.
Clin Chem 1974;20:825-33.
9. Westagrd JO, Biological variation database specification.
The 2010. update. Available at http://www.westgard.com/
biodatabase1.htm. Accessed February 12th, 2011.
10. Billić-Zulle L. Comparison of methods: Passing and Bablock
regression. Biochem Med 2011;21:49-52.
11. van Gammeren AJ, van Gool N, de Groot MJM, Cobbaert
CM. Analytical performance evaluation of the Cobas 6000
analyzer – special emphasis on trueness verification. Clin
Chem Lab Med 2008;46:863-71.
12. Fraser CG, Hytloft Petersen P. Desirable standards for laboratory tests if they are to fulfill medical needs. Clin Chem
1993;39:1447-55.
13. Fraser CG, Hyltoft Petersen P, Ricos C, Haeckel R. Proposed
quality specifications for the imprecision and inaccuracy
of analytical systems in clinical chemistry. Eur J Clin Chem
Clin Biochem 1992;30:311-7.
14. Hyltoft Petersen P, Ricos C, Stöckl D, Libeer JC, Baadenhuijsen H, Fraser C, et al. Proposed guidelines for the internal
quality control of analytical results in the medical laboratory. Eur J Clin Chem Clin Biochem 1996;34:983-99.
15. Ricos C, Perich C, Minchinela J, Alvarez V, Simon M, Biosca
C et al. Application of biological variation – a review. Biochem Med 2009;19:250-9.
Methods performance validation at Cobas c501
16. Westgard JO, Seehafer JJ, Barry PL. European specifications
for imprecision and inaccuracy compared with operating
specifications that assure the quality required by US CLIA
proficiency testing criteria. Clin Chem 1994;40:1228-32.
17. Stöckl D. European specifications for imprecision and inaccuracy compared with US CLIA proficiency-testing criteria.
Clin Chem 1995;41:120-1.
18. Westgard JO, Darcy T. The truth about quality: medical usefulness and analytical reliability of laboratory tests. Clin
Chim Acta 2004;346:3-11.
19. Hyltoft Petersen P, Stöckl D, Blaabjerg O, Pedersen B, Birkemose E, Thienpont L, et al. Graphical interpretation of
analytical data from comparison of a field method with reference method by use of difference plots. Clin Chem
1997;43:2039-46.
20. Challand GS, Li P. The assessment of interpretation of test results in laboratory medicine. Biochem Med 2009;19:146-53.
21. Fraser CG, Hytloft Petersen P. Desirable standards for laboratory tests if they are to fulfill medical needs. Clin Chem
1993;39:1447-55.
22. Westgard JO, Seehafer JJ, Barry PL. Allowable imprecision
for laboratory tests based on clinical and analytical test
outcome criteria. Clin Chem 1994;40:1909-14.
23. Hyltoft Petersen P, Fraser CG. Setting in Clinical Chemistry: Can
competing models based on analytical, biological and clinical
outcomes be harmonized? Clin Chem 1994;40:1865-68.
Validacija metoda za određivanje rutinskih biokemijskih analita na analizatoru
Cobas 6000 model serije c501
Sažetak
Uvod: Cobas 6000 (Roche, Njemačka) je biokemijski analizator za spektrofotometrijska, imunoturbidimetrijska i ion selektivna određivanja koncentracije biokemijskih analita. Ovo istraživanje predstavlja analitičku validaciju s naglaskom na procjenu metoda za rutinsko mjerenje.
Materijali i metode: Validacija je provedena za 30 analita (metaboliti, enzimi, elementi u tragovima, specifični proteini i elektroliti). Istraživanje je uključivalo određivanje nepreciznosti unutar serije (N = 20) i nepreciznost između serija (N = 30), netočnost (N = 30) i usporedbu metoda
sa rutinskim analizatorom (Beckman Coulter AU640) (N = 50). Za validaciju cjelokupnog analitičkog procesa izračunali smo ukupnu pogrješku
(engl. total error, TE). Rezultati su ocijenjeni prema preporučenim kriterijima Radne skupine Europske skupine za procjenu reagensa i analitičkih
sustava u laboratorijskoj medicini (engl. European Group for the evaluation of reagents and analytical systems in laboratory medicine).
Rezultati: Koeficijenti varijacije za nepreciznost unutar serije za sve analite bili su manji od 5%, osim za kolesterol, trigliceride, IgA i IgM. Koeficijenti
varijacije za nepreciznost između serija za sve analite bili su manji od 10%. Sljedeći analiti nisu zadovoljili tražene specifikacije za nepreciznost: ukupni
proteini, albumini, kalcij, natrij, kloridi, imunoglobulini i HDL kolesterol. Analiti koji nisu zadovoljili zahtjeve za netočnost su: ukupni proteini, kalcij,
natrij i kloridi. Analiti koji su odstupali od preporučenih kriterija za ukupnu pogrješku: ukupni proteini, albumini, kalcij, natrij, kloridi i IgM. PassingBablokovom regresijskom analizom dobivena je linearna jednadžba i 95%-ni interval pouzdanosti za odsječak i nagib. Kod određivanja malog broja
analita došlo je do potpunog podudaranja s rezultatima dobivenih na rutinskom analizatoru Beckman Coulter AU640. Kod određivanja drugih analita
postojala je mala proporcionalna i/ili mala stalna razlika u mjerenjima, što znači da je potrebna prilagodba prije uvođenja u rutinski rad.
Zaključak: Temeljem niskih koeficijenta varijacije može se zaključiti da ispitani analizator ima zadovoljavajuću točnost i preciznost te da je izuzetno stabilan. Osim za određivanje analita čiji su rezultati sukladni na oba analizatora, za određivanje nekih analita potrebno je učiniti neke prilagodbe kako bi se postiglo potpuno podudaranje prije uvođenja u rutinski rad.
Ključne riječi: analitička validacija; kemijske analitičke tehnike; analizator Cobas 6000; Passing-Bablockova regresija, ukupna pogrješka
Biochemia Medica 2011;21(2):182-90
190

Download Report